The experimental setup

The test setup for the transmission measurements of a 30 m channel with PAM-32 coding using an arbitrary waveform generator (AWG 70001A from Tektronix).

© Harting

In order to obtain information on the technical feasibility of 100G transmission via symmetrical copper cabling, the task was broken down into the essential criteria to be examined. The core element of 100G data transmission over four pairs of a symmetrical transmission channel is the transmission of 25 Gbit/s over one pair. For further investigation, the complete symmetrical transmission channel for 100G is therefore reduced to its individual components:

• Transmission over 4-pair copper cabling
• Consideration of the cable
• Consideration of the connectors

The individual components can be described mathematically by so-called modeling. These descriptions (matrices) are checked for their accuracy in practice by means of measurements and the mathematically described individual components can later be combined to form an overall transmission channel. This makes it possible to predict limit values, for example with regard to bandwidth (in GHz), near-end crosstalk (NEXT in dB) or length limitation (in meters). This approach made it possible to clearly define the steps required to investigate a 100G channel.

Starting from the PCB channel, the passive transmission channel of the cabling (ISO/IEC channel) could be defined and described completely mathematically using M-matrix data. In this case, the M-matrix is a table with physical properties of connectors, cables and - after calculation - the complete transmission channel. In the broadest sense, the M-matrix therefore stands for the mathematical image/model of the physical elements of a transmission channel and thus represents the link between a real existing cabling system and its physical properties.

NEXT measurements on a plug connection in mated condition. By extending the frequency range up to 2.5 GHz and the continuous development of the values, it can be deduced that these designs are also suitable for higher bandwidths.

© Harting

In order to ultimately fill the M matrices with values, the transmission properties of the individual components need to be determined. To this end, series of measurements were set up for cables and connectors to examine so-called 'mated pairs' - i.e. plug connections that check the plug and socket in the mated state.

To create these measurement series, the measurement equipment of the three partners was completed with the help of the measurement equipment manufacturers and pushed to its limits. At the same time, the measurement setups with connection units, equalizers and DUTs were developed, tested, described and checked for their usability and accuracy at bandwidths of up to 2.5 GHz. Regular round-robin tests were carried out to evaluate the stability and accuracy of the measurement results. Prepared test specimens from all three partners were subjected to the same tests in their laboratories using their respective measurement technology. Specifically, this involved the prototype of a new 2.5 GHz copper cable and existing or appropriately revised connectors. If there were significant deviations in the test results, the causes were analyzed and the measurement setup and test equipment were continuously refined and corrected.

This methodical approach made it possible to comprehensively describe the transmission properties of the individual components. At the end of the test series, reliable values were available for cables and connectors over a frequency range of up to 2.5 GHz - including IL (Insertion Loss), NEXT (Near End Cross Talk), FEXT (Far End Crosstalk), RL (Return Loss), TCL (Transverse Conversion Loss) and ELTCTL (Equal Level Transverse Conversion Transfer Loss). The latter two parameters have been added relatively recently in order to be able to describe (unshielded) UTP cabling in terms of transmission reliability and EMC.

The results of the measurements

The results of the cable measurement: Left at IL - this parameter remains below the calculated limit curve (red). Therefore, IL is fulfilled by the cable. The measurement curve for RL (right) was also above the calculated limit curve.

© Harting

The test results of the various cable prototypes led to a continuous improvement in the design of the PIMF cable construction. PIMF stands for Pair in Metal Foil, a special highly shielded cable construction in which each pair of wires is shielded by a foil (individual pair shielding). The overall construction (all four pairs) is then provided with another overall shield - either in the form of a braided shield or as a combination of foil and braiding. Normally, this type of cable is called S-STP/SF-STP.

A final prototype (pre-series status) was finally able to fulfill all the essential transmission parameters to a largely satisfactory degree. Prototypes of 2.5 GHz cables and connectors were used to construct the transmission channel. These cabling components are electrically characterized by a series of measurable transmission parameters. These include, among others:

• The characteristic impedance (ideally 100 Ω)
• NEXT (near-end crosstalk attenuation)
• IL (attenuation)
• RL (return loss)
• Delay skew (delay time differences)

The test was carried out against an assumed limit curve, which was determined to be necessary for 25 Gbit/s transmission (1 pair).

When investigating the transmission behaviour of connectors, the project partners used existing connector types: the ARJ (specially chambered and highly shielded version of the RJ45), the Tera connector (recognized and standardized Cat.7A connector) and the M12-X-coded connector. This confirmed the assumption that the better the separate shielding of the individual contact pairs in a connector is designed, the better the transmission behavior at high frequencies and large bandwidths.

The reliable test results of the components could then be combined in the channel model. Ultimately, the mathematical investigations based on the model allowed the conclusion that a transmission of 25 Gbit/s over a pair of a symmetrical copper channel of 30 m in length with bandwidths of 2.5 GHz or more is possible. To substantiate the validity of this statement, channel measurements were carried out. The 30 m transmission channel was measured with 26 m of cable and 2 m of preassembled cable (patch cords) using a mobile test device up to 2 GHz. The resulting measurement results confirmed the tests on the model, even if tolerances and inaccuracies are to be expected here.

Further investigations of simplified, 30 m long channels using laboratory measurement technology also provided additional parameters such as IL (insertion loss) and group delay up to 3 GHz - for several pairs simultaneously. All results also support the feasibility of 100G transmission over four pairs of a symmetrical copper channel.

In final tests, the physically constructed transmission channel was finally loaded with useful signals.

The implementation of physical findings

The aim was to find out which coding method makes the most sense for the transmission of 100 Gbit/s via symmetrical copper cabling. The test sequence with a signal sequence based on PAM 16 and PAM 32 was fed in at one end of the purely passive transmission channel and compared with the data received at the other end.

As with the other tests, the first step was again to look at a wire pair - i.e. 25 Gbit/s transmission. Ordered bit sequences (selected sequence of bits) and unordered or stochastic bit sequences (pseudo-random sequence of bits) were routed via the test setup using PAM-16 and PAM-32 coding. The evaluation of the data obtained in this way in the form of so-called eye diagrams led to the conclusion that the PAM-32 coding method should be used for secure data transmission of 100G via a symmetrical copper channel.

The conclusion of the 100G joint project: The technical feasibility of transmitting 100 Gbit/s Ethernet via symmetrical copper cabling systems has been demonstrated. However, the prerequisite is that cables and connectors must comply with the transmission limits described/defined in the project over a bandwidth of 2.5 GHz. Today's symmetrical copper data cables and connectors generally end at 1 GHz (Cat.7A specification). What they are capable of achieving beyond this is usually not really known or tested. We also wanted to explore this in this project.

The next step is to propose the description of the necessary protocols in IEEE 802.3 and the cabling required for this in ISO/IEC JTC 1/SC 25/WG 3.

Authors:
Yvan Engels works in the Strategic Market Development department at Leoni;
Prof. Dr. Albrecht Oehler teaches and researches at the ESB Business School at Reutlingen University;
Rainer Schmidt is Business Development Manager for Industrial Cabling at Harting Electronics.